The dataset at the heart of the study is remarkable in both scale and granularity. Provided by the mobile operator TIM Enterprise, it covers 60 consecutive days in August and September 2019 and reports millions of daily origin–destination flows between Elementary Census Areas, the smallest administrative units in the region, each one enriched with demographic and behavioral attributes. Every record captures how many people moved from one census area to another during a given interval, decomposed into six age cohorts, male and female counts, Italian and foreign citizens, and behavioral labels distinguishing business-related trips from consumer and leisure travel. These attributes are derived from SIM card and contract registration data, while trip purpose is inferred by the operator using proprietary algorithms based on mobility patterns and network usage, not from surveys.

Because raw operator data is subject to privacy-preserving suppression protocols required by the GDPR, many low-volume records contain demographic sub-categories that have been masked or zeroed out. The team therefore applied a stringent multi-step filtering process. They first restricted the analysis to intra-regional movements, then enforced temporal consistency checks that discarded records with negative durations or trips lasting more than an hour, and finally imposed a strict demographic coherence filter, retaining a flow only if the sum of its sub-components across age, gender, nationality, and trip purpose exactly matched the total. This rigorous cleaning reduced the dataset from an initial 36,309,517 raw records to 22,962,575 robust, fully consistent records, discarding more than 13 million heavily censored micro-movements and yielding a dataset of exceptional reliability.

The choice of August and September was deliberate rather than incidental. The two months capture two fundamentally different mobility regimes: August reflects the Italian holiday period, with dispersed, leisure-driven travel, while September marks the resumption of work commutes and the school year. Split violin plots comparing daily mobility distributions between the months revealed exactly this contrast, with August exhibiting lower and highly variable volumes and September displaying higher, tightly concentrated movement patterns. The aggregate time series showed clear weekly seasonality, with pronounced weekend dips, particularly on Sundays, and a gradual upward trend as summer ended. Hourly patterns proved remarkably stable, however: roughly 35 to 40 percent of daily movements consistently occurred between noon and 5 p.m., regardless of the month.

Demographic disaggregation added further texture. People over 60 were consistently the most mobile group during the observation window, likely reflecting seasonal tourism, while minors moved the least. Mobility among foreign citizens remained stable through August before declining in September, consistent with post-summer departures. At the city scale, a case study of Bologna identified the ten most intense internal flows, which clustered around the historic center, the Central Railway Station, and Piazza Maggiore, with strong corridors extending along Via San Donato, Via Andrea Costa, and Via Emilia Levante. Flows between Bologna and its surrounding province radiated outward from the city, confirming its role as the region’s dominant mobility hub, with key connections to Calderara di Reno, Castel Maggiore, San Lazzaro di Savena, and Granarolo dell’Emilia. On weekdays, longer-distance flows concentrated along functional corridors to Modena, Ferrara, and the Romagna coast, while weekend flows spread more evenly toward coastal and recreational areas. During August, the network’s topology remained structurally identical, but absolute volumes on top commuting routes dropped by more than 30 percent.

Beyond description, the researchers built a predictive framework to forecast daily inflows into individual census areas. Feature engineering captured weekly and seasonal cycles through temporal identifiers such as weekday, week of year, and month, along with binary weekend and holiday flags, autoregressive lag features representing arrivals on previous days, and exogenous variables including weather events and localized cultural or sporting events. Regional proxies, namely aggregate inflows from the neighboring hubs of Bologna and Imola recorded the previous day, allowed the model to account for mobility pressure propagating across the network. Target values were log-transformed to handle the strong right-skew of flow distributions, and a Gradient Boosting Regressor, optimized via grid search, was trained on 70 percent of the data and tested on the remainder. Urban and industrial hubs such as Bologna and Modena proved the most predictable, with errors of roughly 10 to 15 percent attributable to the stability of commuting patterns, while tourist-centric and mountainous areas showed errors below 25 percent, remaining within bounds suitable for planning purposes even in highly variable zones.

Perhaps the most striking demonstration, however, is the automated event detection. For each census area and hour, the team computed a Z-score comparing the observed net flow against the historical mean and standard deviation for that location. Values exceeding three standard deviations in either direction were flagged as potential anomalies. The method worked spectacularly well where the signal-to-noise ratio was high. At the census area containing the Bologna Exhibition Centre, a pronounced anomaly from 23 to 27 September aligned precisely with the opening of Cersaie, the international ceramics fair that draws tens of thousands of visitors, while a weaker peak in early September corresponded to the SANA organic products exhibition. At the Imola Autodrome, anomalies on 7 September matched the CRAME Swap Meet, one of Europe’s largest vintage vehicle fairs, complete with the characteristic outbound spike at 5 p.m. as crowds dispersed, and a further cluster from 20 to 22 September captured the Summer Food Experience. In Modena, Z-scores exceeding 3.5 tracked the start of the school year, the Festivalfilosofia festival, and the Modena Motor Gallery.

The study is unusually honest about the method’s limits. Dedicated event venues show near-zero baseline mobility on quiet days, producing a low standard deviation and a very high signal-to-noise ratio, so events emerge cleanly. Mixed-use urban centers, by contrast, carry heavy routine traffic from commuters, students, and residents, inflating the historical standard deviation and dampening Z-scores, which caused mid-sized events like the SANA exhibition to barely reach the detection threshold. Conversely, urban areas are prone to false positives: a massive anomaly in Modena on 30 September, with a Z-score of 5.91, turned out to be ordinary end-of-month congestion combined with weather-related disruptions rather than any public event. The authors propose context-aware baselines, comparing, for instance, a Tuesday evening only against historical Tuesday evenings, as a future remedy.

The final phase of the work zoomed in on Dozza, a small historic tourist town near Bologna whose limited number of access points makes it ideal for systematic monitoring. Smart cameras installed at three locations recorded hourly pedestrian entries and exits from March 2022 to September 2025, with measurements correlating at better than 0.95 across cameras. Entries peaked around midday and exits around 5 p.m., with weekends showing substantially higher footfall. The team fused this ground truth with a TIM presence dataset that classifies mobile users as tourists, excursionists, transients, or residents. Only the excursionist category correlated significantly with camera counts. Crucially, the analysis revealed that pedestrian peaks in Dozza were often driven by events beyond the municipality itself, including the Formula 1 Grand Prix at Imola, major trade fairs such as MECSPE, COSMOPROF, and CERSAIE in Bologna, and activity surges at Bologna Airport, effects the researchers captured through simple lagged binary indicators marking weekends following high-activity weeks.

Two complementary Gradient Boosting experiments, evaluated with leave-one-month-out cross-validation, showed that the mobility-based feature set achieved better absolute accuracy, with an RMSE of 218.26 and MAE of 152.02, while the event-based set achieved a lower MAPE of 33.40 percent and, critically, predicted the exact dates of pedestrian peaks more reliably. A probabilistic classifier identifying peak days above 500 pedestrians flagged at least one probability peak in every observed high-attendance period. These models are now embedded in an operational decision-support application built for the municipality, structured around a backend service, a web frontend, a PostgreSQL database, and a pedestrian prediction service, allowing administrators to define what-if scenarios for parking and public transport and to see the estimated infrastructure load under low-event and high-event conditions. The researchers argue that the framework offers a dual-purpose tool, supporting both regional transport planning and fine-grained tourism management, and that future work will extend it to real-time data sources and multi-modal, regional-scale scenarios, bringing data-driven sustainability a step closer to everyday urban governance.

The study, authored by Ricardo Prieto-Curiel, leverages an extensive dataset curated by CitiesMoving.com, a comprehensive platform that harmonizes mobility surveys from cities worldwide according to the ABC framework. This framework classifies journeys into three key categories: active mobility (A), which includes walking and cycling; bus and other forms of public transport (B); and private car use (C). By examining modal shares across various European locales, the research distinguishes between cities equipped with metro systems, tram-only cities, and those lacking any rail infrastructure. The findings present a nuanced narrative about how these different transportation frameworks influence urban travel behavior—particularly the extent to which they succeed in mitigating car use.

One of the study’s pivotal insights is the marked difference in car usage between cities possessing metro systems and those with only tram networks or none at all. Metro-equipped cities demonstrate a considerably lower share of journeys made by car. Such a trend suggests that rapid transit systems, with their capacity for high-frequency, long-distance travel, more effectively dissuade private vehicle use compared to trams, which often have lower capacities and shorter ranges. This observation not only challenges common perceptions that all rail transit equally curbs car dependency but also underscores the unique role of metro systems within the urban transport ecosystem.

The distinction between metros and trams lies not only in their physical infrastructure but also in their operational frameworks and network integration. Metros typically run on dedicated tracks, separated physically from other traffic, ensuring consistent speeds and reliability. Their ability to cover greater distances without frequent stops and their incorporation into larger multimodal networks make them a dominant choice for commuters traveling longer distances within metropolitan areas. In contrast, trams often share road space with other vehicles, which can subject them to delays caused by traffic congestion. Their shorter routes and more frequent stops make them better suited for localized travel, potentially limiting their effectiveness in replacing car trips.

Cost and accessibility factors are also instrumental in explaining why metro systems significantly reduce car journeys while trams do not show the same impact. Construction and operational expenses for metros are considerably higher, confining their presence to economically robust cities or those with large population densities justifying such investment. This exclusivity might mean that residents in metro cities have fewer viable alternatives to private cars, making the metro’s efficient service a more attractive option. Tram systems, though generally less costly to deploy and maintain, primarily serve smaller urban areas or act as feeders to other transit modes, which may diminish their overall influence on curtailing car use.

The health, economic, and environmental benefits of active mobility and public transportation are well-documented, rendering the insights of this study particularly salient. Cities are increasingly championing policies designed to encourage walking, cycling, and the use of public transit to mitigate the environmental footprint of urban travel. Yet, this research provides a sobering reminder that not all public transport solutions yield equal returns in these efforts. Policymakers must therefore evaluate the specific context of their cities, considering whether investments in metro infrastructure might yield more substantial reductions in automobile dependency than expansions of tram networks or other forms of public transit.

Moreover, the study’s utilization of the ABC framework adds a layer of sophistication by allowing analysts to harmonize data across diverse cities and compare transport modes on a standardized basis. This methodological approach addresses the common challenge of inconsistent data collection methodologies that has often hindered cross-city comparisons in mobility research. By categorizing journeys via active mobility, bus/public transport, and private cars, the framework provides a clear lens through which the complex interplay of urban transport modes can be understood, facilitating actionable insights.

Another noteworthy aspect of the study is the geographic focus on European cities. Europe’s extensive diversity in urban morphology, economic development, and transport infrastructure offers fertile ground for such comparative analyses. From sprawling metropolises with advanced metro networks to smaller cities relying primarily on trams or buses, the continent embodies a spectrum of structural and cultural approaches to urban mobility. The research findings thus carry important implications beyond Europe, shedding light on transit planning strategies that could be applicable to emerging cities worldwide grappling with similar challenges.

Environmental imperatives add urgency to these transportation debates. Private vehicles are among the largest contributors to urban air pollution and greenhouse gas emissions, exacerbating climate change and harming public health. The capability of metro systems to lower car usage directly translates to decreased emissions, less noise pollution, and improved air quality. By illustrating this link empirically, the study galvanizes support for the expansion of metro infrastructure as an integral component of green urban policy strategies. Conversely, the limited impact of trams on car reduction signals that tram investments, while beneficial for other aspects of urban transport, should be complemented by other initiatives to maximize environmental benefits.

The socio-economic dimensions of the study’s findings should not be overlooked. The reduction in car dependency facilitated by metro access often correlates with enhanced social equity. Metro systems typically serve a broad cross-section of urban residents, enabling affordable access to jobs, education, and amenities without the need for private vehicle ownership. This can alleviate economic burdens on low-income populations and increase overall urban inclusivity. The research thereby highlights a compelling social justice argument in favor of metro development, which aligns with broader goals of creating livable, equitable cities.

Technological advancements, such as real-time transit tracking, integrated fare systems, and electrification of fleets, further enhance the appeal and efficiency of metro networks. These innovations contribute to a seamless passenger experience that can draw travelers away from the convenience of private car use. While tram systems can also benefit from such technologies, their inherent operational constraints—like vulnerability to street-level conditions—may limit the extent to which these technologies can transform their effectiveness relative to metros.

The study also invites reflection on the future trajectories of urban transport amidst evolving lifestyles and work habits. The COVID-19 pandemic fundamentally altered commuting patterns, with increases in remote and hybrid work reducing overall transit ridership temporarily. However, as cities adapt to post-pandemic realities, the role of reliable, efficient public transport remains pivotal. Investing in metro systems can offer a resilient backbone for urban mobility, adapting dynamically to fluctuating demand while continuing to discourage excessive reliance on private vehicles.

Importantly, the findings challenge urban planners and policymakers to critically evaluate incremental improvements to tram systems versus transformative investments in metro expansions. While cost considerations often favor tram enhancements, these may not translate into meaningful reductions in car use without complementary policies such as congestion pricing, improved pedestrian infrastructure, and multimodal integration. A holistic approach that views metro development within a broader ecosystem of sustainable transport initiatives is essential.

In conclusion, the comparative analysis conducted by Prieto-Curiel provides compelling empirical evidence that metro systems hold a unique and potent capacity to reduce car dependency in European cities, outperforming trams and cities without rail systems. This revelation holds profound implications for urban mobility planning, environmental sustainability, and social equity. As cities worldwide face escalating pressures to decarbonize and improve quality of life, prioritizing the development and expansion of metro networks could be one of the most effective strategies to achieve these goals. This study, therefore, serves as a clarion call to rethink urban transport investments through the lens of impact, scalability, and long-term benefits.

Subject of Research: The study investigates the impact of different rail-based public transport systems—metros and trams—on private car usage and mobility patterns in European cities.

Article Title: Metros reduce car use in European cities but trams do not

Article References:
Prieto-Curiel, R. Metros reduce car use in European cities but trams do not. Nat Cities (2025). https://doi.org/10.1038/s44284-025-00342-7

DOI: https://doi.org/10.1038/s44284-025-00342-7